Almost everyone seems to be attempting to get a bit of the generative AI motion nowadays. While most of the main focus stays on model providers like OpenAI, Anthropic, and Cohere, or the larger corporations like Microsoft, Meta, Google, and Amazon, there are literally many startups attempting to tackle the issue of generative AI in quite a lot of ways.
Fireworks.ai is one such startup. While it lacks the brand recognition of a few of these other providers, it has the most important open source model API with over 12,000 users, in response to the corporate. This form of open source motion tends to draw investor attention, and the corporate has raised $25 million to this point.
Lin Qiao, co-founder and CEO of Fireworks, points out that her company doesn't train basic models from scratch, but reasonably helps tailor other models to an organization's specific needs. “It might be either standard models, open source models, or models that we optimize or models that our customers can optimize themselves. All three variants might be deployed via our inference engine API,” Qiao told TechCrunch.
Because it’s an API, developers can integrate it into their application, train the model of their selection on their data, and add generative AI features comparable to: B. ask questions in a short time. According to Qiao, it’s fast, efficient and delivers high-quality results.
Another advantage of the Firework approach is that it allows corporations to experiment with multiple models, which is essential in a rapidly changing market. “Our philosophy here is that we wish to provide users the flexibility to iterate and experiment with multiple models and have effective tools to feed their data into multiple models and test against one product,” she said.
Perhaps more importantly, they keep costs down by capping the model size at 7 to 13 billion tokens, in comparison with over 1 trillion tokens for ChatGPT4. While this limits the range of words the big language model can understand, it allows developers to concentrate on much smaller, focused data sets designed to work with more limited business use cases.
Qiao is uniquely qualified to construct such a system, having previously worked at Meta, where he led the AI platform development team with the goal of constructing a rapid, scalable development engine to power AI across all services to advance meta. She was capable of take this data from her work at Meta and develop an API-based tool that brings this power to any company without requiring the technical resources of an organization the scale of Meta.
The company raised $25 million in 2022 led by Benchmark with participation from Sequoia Capital and angel investors including Databricks and Snowflake. The latter two are particularly interesting strategic investors because each are data storage tools and Fireworks will enable users to leverage that data.
